Next-generation sequencing of paired tyrosine kinase inhibitor-sensitive and -resistant EGFR mutant lung cancer cell lines identifies spectrum of DNA changes associated with drug resistance

Abstract

Somatic mutations in genes encoding kinases are associated with increased sensitivity of some solid tumors to kinase inhibitors, but patients with metastatic cancer eventually develop disease progression. A common method used to model acquired resistance involves culturing parental drug-sensitive cells with increasing concentrations of drug until cells emerge that are resistant. In EGFR mutant lung cancer, this modeling has reliably identified clinically relevant EGFR tyrosine kinase inhibitor (TKI) resistance mechanisms such as the second-site mutation, EGFR T790M, amplification of the gene encoding an alternative kinase, MET, and epithelial-mesenchymal transition (EMT). The full spectrum of DNA changes associated with EGFR TKI acquired resistance remains unknown. Here, we used next-generation sequencing and bioinformatics analysis to characterize mutational changes associated with 4 populations of EGFR mutant drug-sensitive cell lines and 5 matched drug-resistant cell lines. Comparing resistant cells with their parental counterparts, we identified 16-89 coding SNVs/indels that were acquired and 1-27 that were lost; few SNVs/indels were shared across resistant lines. Comparison of two related parental lines revealed no unique coding SNVs/indels, suggesting that the changes in the resistant lines were due to drug selection. When analyzing whole genome sequencing data from one isogenic pair, we found that there was a higher frequency of SNVs in 'constant late' replication timing zones as compared to 'constant early' replication timing zones (chi-squared p-value < 10-5) and an enrichment of SNV frequencies in genomic regions harboring lamina-associated domains compared to the remainder of the nucleus (chi-squared p-value < 10-5). Surprisingly, we observed a higher burden of CNV changes across all resistant lines, and the one line that had an EMT phenotype displayed significantly higher levels of CNV changes than the other lines with acquired resistance. These results demonstrate a framework for studying the evolution of drug-related genetic variants over time and provide the first genome-wide spectrum of mutations associated with the development of cellular drug resistance in an oncogene-addicted cancer. Collectively, the data suggest that CNV changes may play a larger role than previously appreciated in the acquisition of drug resistance and highlight that resistance may be heterogeneous in the context of different tumor cell backgrounds.